How to download and use nv_dinov2 in tao dino object detection?

I tried to download like this "ngc registry model download-version nvidia/tao/pretrained_dino_imagenet:vit_large_nvdinov2 --dest “$LOCAL_PROJECT_DIR/dino/” but its not working could you please tell me how to use this model and back bone for training i am using tao docker version 5.3.

May I know the log?

You can refer to model card TAO Pretrained DINO with Foundational Model Backbone | NVIDIA NGC and its file TAO Pretrained DINO with Foundational Model Backbone | NVIDIA NGC.

You can run the dino notebook.

total 24
drwxrwxr-x 2 taotoolkituser taotoolkituser 104 Aug 14 2023 annotations
drwxrwxr-x 7 taotoolkituser taotoolkituser 16384 Aug 11 2023 train2017
drwxrwxr-x 2 taotoolkituser taotoolkituser 4096 Aug 11 2023 val2017
drwxrwxr-x 2 taotoolkituser taotoolkituser 6 Aug 11 2023 validation
–2024-04-02 03:53:02-- https://ngc.nvidia.com/downloads/ngccli_cat_linux.zip
Resolving ngc.nvidia.com (ngc.nvidia.com)… 13.35.7.87, 13.35.7.68, 13.35.7.106, …
Connecting to ngc.nvidia.com (ngc.nvidia.com)|13.35.7.87|:443… connected.
HTTP request sent, awaiting response… 200 OK
Length: 47835508 (46M) [application/zip]
Saving to: ‘/opt/nvidia/tools/tao-experiments/ngccli/ngccli_cat_linux.zip’

ngccli_cat_linux.zip 100%[===============================================================================================================>] 45.62M 30.3MB/s in 1.5s

2024-04-02 03:53:04 (30.3 MB/s) - ‘/opt/nvidia/tools/tao-experiments/ngccli/ngccli_cat_linux.zip’ saved [47835508/47835508]

Archive: /opt/nvidia/tools/tao-experiments/ngccli/ngccli_cat_linux.zip
listing the models
‘NoneType’ object is not iterable
Error: ‘nvidia/tao/pretrained_dino_imagenet:vit_large_nvdinov2’ could not be found.
Check that model is downloaded into dir.
ls: cannot access ‘/opt/nvidia/tools/tao-experiments/dino/pretrained_dino_nvimagenet_vvit_large_nvdinov2 /’: No such file or directory

I already ran the notebook it works fine other models but nv_dinov2 is not working the path i am trying to download the pretrained model is not correct its seems.

May I know where did you find above command? I did not see it in tao_tutorials/notebooks/tao_launcher_starter_kit/dino/dino.ipynb at main · NVIDIA/tao_tutorials · GitHub.
Also, from Pre-trained DINO NvImageNet weights | NVIDIA NGC, there is not nv_dinov2 model in Pre-trained DINO NvImageNet weights | NVIDIA NGC.

We provide nv_dinov2 in another model card , TAO Pretrained DINO with Foundational Model Backbone | NVIDIA NGC.

I directly downloaded the model from the link that you gave its working fine now thank you so much @Morganh

Great. Thanks for the info.